high-definition ndvi sampling of arabian protected areas...

High-Definition NDVI Sampling of Arabian Protected Areas to Assess Rangeland ConditionDr David Gallacher

Zayed University

Tamer Khafaga MSc

Dubai Desert Conservation Reserve

What’s the problem?

• Proof of vegetation change is difficult• Historical data is lacking in Arabia• Large spatiotemporal variation

• Rangeland is 70% of world’s land surface

• Most important is• Species ratios• Biomass trends

Red, red edge, infrared, or RGB?

NDVI• Developed in 1970s to interpret satellite data• Proxy measure of chlorophyll (related to growth)• Broad scale: predict NPP (regional crop yields, droughts, forestry

status)• Narrow scale: agriculture (spatial analysis of monoculture)

Green / RGB• Proxy measure of palatable biomass?

Red / Red edge• Proxy measure of wood?

Red, red edge, infrared, or RGB?

Affected by• Plant health

• Hydration & wind (leaf angle)

• Dust

• Phenology

Higher resolution means• More detail (good!)

• More heterogeneity (tricky!)

• Therefore, analysis changes with scale

Plants & Pixels

Vegetation index values low, due to

• Sparse plants

• Low chlorophyll

• Low within-canopy leaf area index

Values for non-plant matter

• Sand

• Gravel

• Water

Questions

• Do vegetation indices have value?

• Are values affected by substrate / aspect / shadow?

• Are background values consistent?

NormalisedDifference Vegetation IndexNDVI =

𝑁𝐼𝑅 −𝑅𝑒𝑑

𝑁𝐼𝑅+𝑅𝑒𝑑

Red edge chlorophyll index

RECI =𝑁𝐼𝑅

𝑅𝑒𝑑𝐸𝑑𝑔𝑒− 1

Results

• Do vegetation indices have value?

• Are values affected by substrate / aspect / shadow?

• Are background values consistent?

Results

• Do vegetation indices have value?

Yes, but need further work

• Are values affected by substrate / aspect / shadow?

• Are background values consistent?

Results

• Do vegetation indices have value?

Yes, but need further work

Maybe no better than RGB

• Are values affected by substrate / aspect / shadow?

• Are background values consistent?

Results

• Do vegetation indices have value?

Yes, but need further work

Maybe no better than RGB

• Are values affected by substrate / aspect / shadow?

• Are background values consistent?

Results

• Do vegetation indices have value?

Yes, but need further work

Maybe no better than RGB

• Are values affected by substrate / aspect / shadow?

• Are background values consistent?

Results

• Do vegetation indices have value?

Yes, but need further work

Maybe no better than RGB

• Are values affected by substrate / aspect / shadow?

Yes

• Are background values consistent?

Results

• Do vegetation indices have value?

Yes, but need further work

Maybe no better than RGB

• Are values affected by substrate / aspect / shadow?

Yes

But indices without RGB are better

• Are background values consistent?

Results

• Do vegetation indices have value?

Yes, but need further work

Maybe no better than RGB

• Are values affected by substrate / aspect / shadow?

Yes

But indices without RGB are better

• Are background values consistent?

No

Results

• Do vegetation indices have value?

Yes, but need further work

Maybe no better than RGB

• Are values affected by substrate / aspect / shadow?

Yes

But indices without RGB are better

• Are background values consistent?

No

Need algorithms to calibrate results

Conclusion

• Use other vegetation indices (not NDVI)

• Need algorithms to interpret biomass proxy

• For many purposes, RGB is equal or better